CN115667632A - Method for automatically and autonomously controlling a tamping machine - Google Patents

Abstract

A method for automatically autonomously controlling a tamping machine (C) having a distance measuring device (WMS, GPS, 32) for precisely detecting the position of a track-laying working machine in a track, and having a signal detection device for an actuator of a working unit (23, bv, 18, 26) of the tamping machine (C) is described. In order to advantageously achieve an automatable tamping action, it is proposed that, during tamping, ballast bed data be detected by means of sensors (23, bv, 18, 26) and current ballast bed parameters be obtained therefrom and stored for a subsequent work process and analyzed by means of a machine learning device (17, ML), wherein an analysis of ballast bed condition data (EF 7, S9, A3) is created on the basis of a machine learning technique (ML, 17), and the ballast bed parameters are analyzed while taking into account tamping force disturbances occurring in the longitudinal direction of the track and a plurality of work instructions (EF 7, S9, A3) for an optimal work mode are determined therefrom and stored, wherein, during the subsequent work process, the tamping machine executes these work instruction data fully automatically and autonomously as a function of the current position in the track and the relevant work instruction data.

Description

Method for automatically and autonomously controlling a tamping machine

Technical Field

The invention relates to a method for automatically and autonomously controlling an upper construction machine on a track, which has a travel measuring device and a precise synchronization device relative to the track, a position detection device of a working unit of a tamping machine, by means of which position detection device a position-precise working instruction for each sleeper area to be tamped is transmitted to a control computer of the tamping machine, and the tamping machine executes these working instruction data fully automatically and autonomously as a function of the current position in the track and the associated working instruction data.

Background

Most railway tracks are designed as ballast tracks. In this case, the sleeper lies flat in the ballast. Due to the wheel forces of the train travelling over it, the ballast is rounded off, partially broken and worn. This leads to irregular subsidence in the ballast and to a geometrical shift in the lateral and vertical track position. The disadvantages of longitudinal section height, superelevation (in curves), distortion, gauge and orientation position occur due to the sinking of the ballast bed. These defects in turn increase the force, which in turn has a destructive effect on the ballast.

If these geometries exceed certain comfort or safety limits specified by the railway administration, maintenance work is planned and carried out in time. Today, upper line engineering machines are often used in order to eliminate and correct these geometrical track defects. To control the process there is a measurement system for detecting the current track position for parameter correction, lift, twist and lateral tilt. In order to be able to put the rails into operation again after such maintenance work, the work machines for the topside of the track are equipped with a so-called acceptance measuring system or an acceptance recorder system. The railroad authority has specified so-called acceptance tolerances for the quality of the track location after mechanical or other improvement by the works on the top of the line. Acceptance tolerances represent the minimum requirement for the quality of the geometric improvement produced.

In addition to acceptance tolerances, there are also safety-relevant open tolerances. These tolerances represent the limits that must be observed so that the treated track can be opened again for train traffic without risk. Compliance with these tolerances is demonstrated by an acceptance recorder system. On the construction machine (tamping machine) located on the top of the track, there is a front driver (vorwagenfuhrer) who is responsible for controlling the machine in terms of nominal geometry and in terms of the track position left after the maintenance work is recorded by the recorder system.

Currently, track maintenance is based on a track geometry plan, which is detected by the position of the rails. The track measuring vehicles travel at regular distances on the track and detect the geometric position of the track. In this case, the track position is generally divided into sections of about 200m long, and the height position, direction, superelevation, and standard deviation of twist are detected. In addition to these statistics, individual single term errors are measured. If the statistical value exceeds a certain comfort tolerance, maintenance work is planned and performed. If the individual errors exceed a certain threshold value, action is taken directly and the individual errors are eliminated immediately, since otherwise a slow-moving position or a track lock must be carried out due to a hazard to train traffic.

The maintenance machine gets the track defects as a preset track nominal geometry, previously recorded and measured and the area that should be maintained. No additional presets are made.

A second operator, the tamping worker, is provided for the tamping process. Regardless of the type of defect in the ballast, he usually performs standard tamping. Whether tamping is carried out for multiple times, position is excessively improved and the like depends on the method.

The tamping worker's estimates, experience and motivation are followed as to exactly how the tamping worker tamps the track, e.g. what tamping pressure he uses, what carriage time, whether he tamps the same sleeper multiple times, whether he handles one location with slightly more lift, whether he chooses slightly larger opening widths, etc.

Before using the work machine on the line, the track position is measured with various known measuring systems and compared with the nominal track position. The difference in height and orientation is transferred as a track correction value to the tamping machine together with the track nominal geometry.

There are tamping machines which are used exclusively for tamping points (divisible tamping units, so-called split-head units, additional lifting devices for turning lines, pivotable lifting hoes, etc.) and tamping machines which are preferably manufactured for tunnel tamping. Tamping machines are known to run with periodic or continuous job priorities. In addition, single-sleeper and multi-sleeper tamping machines are also known. A multi-sleeper tamping machine tamps a plurality of sleepers at one time during an operating cycle. However, multi-sleeper tamping machines may also be used in which only one sleeper is to be tamped.

Today, construction site planning is based on track geometry measured with track measuring vehicles, but irrespective of what causes the track defects, such as ballast in the rail joint area may be damaged, rounded and crushed, similar defects occurring in very hard foundations and being referred to as "white spots". In these positions, the driving dynamics lead to the crushed stone ballast being crushed and these positions are indicated by the escaping mineral dust. The ballast can be very damaged in the case of a long service life of the pavement. A large proportion of fine constituents and organic material or soil pressed up from the foundation may already fill the interstices of the ballast. It is known from practice that the track position in such ballast constructions cannot be permanently corrected with track-tamping machines. It is also known from practice that the individual errors occur randomly scattered in the track. Approximately 40% of these partial fault locations can be permanently eliminated. 60% of these defects develop again in a short time. A track with good ballast conditions is tamped approximately every four years on average. A single error indicating a broken ballast requires maintenance measures to be taken approximately every 1-3 months. In each tamping operation, the tamping tool damages the ballast of the ballast due to the high tamping force. The use of tamping machines results in costs for maintenance work and hinders operation. The train cannot travel during operation and a crawl position is set for adjacent tracks if they are in use. It is also known that highly contaminated ballast beds require high compaction forces. There are virtually no moving gaps between the ballast bed particles, since they are filled with fine material. This increases the tamping force that must be applied to move the ballast. At the same time, the durability of the corrected track geometry of such contaminated road beds is reduced, because the friction and the engagement between the ballast bed particles are smaller. The track geometry defects are usually recorded, stored and transmitted electronically to the computer of the tamping machine by a separate measuring process prior to the tamping operation. The track geometry defects typically have a wavelength of 10-25m and an amplitude of 10-40mm. Long-wave defects in the range of 25-70m also occur, which have a higher defect amplitude.

The tamping unit fixes the position of the track during maintenance measures. This takes place by means of a tamping tool, a so-called pick, which is inserted into the ballast next to the sleeper and tamps the ballast below the sleeper by means of a linear closing movement superimposed by the tamping vibration. In this case, the linear closing movement is superimposed by the hydraulic cylinder and the vibration amplitude produced by the eccentric shaft mechanism, according to the standard. Newer fully hydraulic tamping drives produce both linear closing motion and vibration.

A line upper construction machine is known from WO 2019091681 A1, which detects network data and transmits the network data to a system control center. The work machine on the line has a sensor system and collects raw data. It should thus be planned when and where the work machine on top of the line is to be used. During the navigation process, the raw data is detected in order to update the network data, i.e. data such as rebuilds or faults, instead of the specific ballast parameters detected during tamping. The trend of the variation of the compaction force cannot be obtained from the collection of network data. For example, AT 513973A discloses a fully hydraulic drive for a tamping unit. For the regulation and control of the drive, the adjustment movement is detected by an integrated path sensor. The tamping pressure is measured here by a pressure sensor. As described in AT 520117a, parameters such as tamping work, ballast hardness, ballast bed contamination, tamping force, tamping time and ballast stiffness can thus be measured and derived. It is known from AT 515801a how to specify an optimum tamping time based on measurements. The opening width of the tamping tool can also be freely and continuously adjusted by means of a fully hydraulic tamping drive.

Currently, the tamping operator selects the correct setting of the tamping unit, such as tamping pressure, carriage delivery time, lowering speed of the tamping unit, opening width, tamping depth, lifting of the track or multiple tamping operations, etc. No further work planning, such as tamping work itself, nor preparation work, such as ballast replacement in the region of the local fault location, local drainage improvement, etc., is carried out. This increases the track maintenance costs and reduces the durability of the obtained track position.

The locations in the track having high bed stiffness constitute high points whose height position does not vary much due to train traffic. The more the stiffness fluctuations in the subgrade (or track bed) are different, the greater the force interaction between the wheels and the rails, the higher the load of the track and the faster the track geometry deteriorates. Individual short fault locations in the rail have a tendency to extend in the longitudinal direction under the action of high dynamic forces in the rail, increase the height of the rail defect and produce subsequent defects caused by the excited rail vehicle.

Known sensors can determine the position of ties in a track as a tamper passes over the ties. By means of this device, the machine can be positioned correctly for tamping in a fully automatic manner. Fully automatic working machines are therefore known in practice.

In some cases, the track geometry rating data is stored in a database of the infrastructure operator and the data can be downloaded or the results can be partly returned. Machine learning systems are prior art. Machine learning is a generic concept of computer-aided knowledge generation from experience. To this end, the algorithm builds a statistical model based on the training data. Patterns and regularity are recognized in the learning data. Thus, the system can also evaluate unknown data. The sleeper and the recorded measurement parameters can be assigned precisely to the track kilometers by means of GPS coordinates by means of a GPS system installed on the tamping machine.

A virtual GPS correction data service is known which sends correction data to the appropriate GPS receiver. Only one measuring vehicle moving on the rails supported by the moving GPS is thus required. The advantage of RTK (real time kinematic) -GPS is that it can determine the absolute position (position about 5mm, height about 10-15 mm) very accurately by means of RTK-correction data. The more satellites and satellite systems that are simultaneously received by the GPS receiver, the more accurate the results. Modern satellite receivers receive and utilize the satellite systems GPS, GLONASS, GALILEO, beidou, QZSS, IRNSS and SBAS simultaneously. These satellite systems may send data to the correction service and receive correction data on a second channel. An accuracy in the range of 5-15mm is too inaccurate for calculating correction values for the track tamping machine for lifting or straightening, but this accuracy is sufficient to determine an absolute reference point for the track geometry. Sleepers in the track and in other locations can also be positioned equally precisely and provided with GPS coordinates. These positions or sleepers can be retrieved accurately and uniquely by means of an over-the-line engineering machine equipped with an RTK-GPS system.

Disclosure of Invention

The object of the invention is to provide a method for automatically and autonomously controlling a construction machine on a line, which avoids the above-mentioned disadvantages. The method is intended not only to provide the engineering machines on the track with the nominal geometric data and the track position correction data in general, but also to provide precise, locally uniquely assigned work instructions in order to achieve autonomous tamping with high quality and in accordance with the properties and requirements of the ballast bed, thereby avoiding human error-prone. At the same time, ballast bed parameters are detected by the tamping machine during the operation, analyzed by the computer and preferably transmitted to the infrastructure operator after the end of the operation in order to prepare it for the next overhaul.

The invention solves this technical problem by the features of claim 1. Advantageous further embodiments of the invention are described in the dependent claims. The invention solves the technical problem, inter alia, by detecting ballast bed data by means of sensors during tamping and acquiring current ballast bed parameters therefrom and storing them for a subsequent work process and analyzing them by means of a machine learning device, wherein an analysis of the ballast bed situation data is created on the basis of a machine learning method and the ballast bed parameters are analyzed taking into account a tamping force disturbance occurring in the longitudinal direction of the track and a plurality of work orders for an optimal work process are determined therefrom and stored, wherein the tamping machine executes these work order data fully automatically and autonomously as a function of the current position in the track and the relevant work order data during the subsequent work process.

In order to automatically and autonomously control the operation of the tamping machine and its tamping and lifting straightening units, the following operations are carried out: the control computer of the tamping machine is supplied with positionally accurate operating instructions (for example, by means of GPS coordinates) for each sleeper area to be tamped (this may include multiple tamping operations, a greater opening width of the tamping tool, tamping pressure, excessive lifting, presetting of the maximum tamping force, tamping time, automatic tamping time according to the tamping force or the like, or presetting of the operating sequence in the switch, for example, at which positions the split-head unit is split in the switch tamping operation and the outer part is pivoted outward, etc.). These operating parameters have been detected in a previous operating process, i.e. in a complete or partial tamping of the track, and have been stored for a subsequent operating process.

The tamping machine is precisely positioned on the area of the sleepers to be tamped by automatic sleeper recognition or GPS coordinates. In the reached position, the tamping machine can then execute these work orders fully automatically and autonomously according to the preset work orders, in the process of which, if necessary, new work orders are generated for the next process and then moved by the automatic travel system to the next sleeper area, where the process is repeated accordingly until the entire predetermined work area has been processed.

The predetermined work instructions do not have to be executed fully automatically, but can be displayed for each tie area to the operator, who sets and implements the predetermined work mode.

According to the invention, during tamping, ballast bed data and operating data are detected by means of the fully hydraulic tamping drive and its sensors, and current ballast bed parameters (such as ballast bed hardness, tamping force, tamping time, immersion time of the tamping unit, braking acceleration of the tamping unit during immersion, current GPS position or track number, current lifting and correcting value, current lifting and correcting force, etc.) are calculated and stored therefrom, and are analyzed by means of machine learning techniques by means of a machine learning device.

According to the invention, a ballast bed condition record is created during the operation and displayed to the tamping worker or the front truck operator for him to learn, and a ballast bed condition report is generated from the measurement data after the operation, wherein the record and the report are both sent to the infrastructure operator as a basis for the preparation of the operation for the next tamping overhaul.

If the infrastructure operator or responsible foreman does not transmit work instructions, the analysis of the ballast bed data carried out therewith during the tamping operation is transmitted to the tamping worker for prompting the optimal work mode.

In accordance with the invention, the measurement data of the tamping operation are analyzed by a rule-based expert system (AI system or other machine learning program) with regard to sudden tamping force disturbances (individual defects) in the longitudinal direction or statistical parameters such as standard deviation, mean values, etc. in relation to rail position defects, etc., and the instructions for an optimal operating mode are derived and preset therefrom.

Useful and valuable information can be obtained from the data determined by the hydraulic tamping drive and its sensors. The analysis is performed using algorithms from the field of Artificial Intelligence (AI). Artificial intelligence systems are able to find correlations and patterns in data sets of different structures that are hardly or not at all artificially detectable. By means of the AI system, predictions about the deterioration of the track position and the occurrence of track defects can be created and maintenance recommendations for improving the track position durability can be made therefrom. Other Machine Learning (ML) techniques (rule-based learning) are also suitable for this purpose.

The rule-based expert system (XPS) may assist the operator with specific recommendations. XPS has the major advantage in the area where there is a profound indication for interpretation of the algorithm model and data layers.

The following exemplarily illustrates possible forms of job instructions. The job instruction may be generated by a job master computer. The table here includes all the sleepers to be tamped.

The various job instructions may be agreed upon and standardized in coordination between infrastructure operators and machine operators. Here, the meanings of these job instructions are:

EF 7-single defect in the track. The operation parameters are as follows: tamping for three times; expanded opening width (75 cm); tamping time 3x 0.65s

S9-highly contaminated crushed stone ballast area. The operation parameters are as follows: tamping for one time; tamping pressure 180 bar; tamping time 1.2s

A3-general Preset. The operation parameters are as follows: the maximum tamping force is limited to 35kN; tamping with automatic optimized tamping time selection; tamping depth 485mm

By prescribing in advance general presets (working parameters which should be used when there are no specific requirements) and working instructions for a specific location and their nomenclature, the control computer of modern tamping machines readily interprets and executes the correspondingly defined instructions.

Drawings

The technical solution of the invention is exemplarily shown in the drawings. In the drawings:

fig. 1 shows a schematic side view of a tamping machine;

fig. 2 shows a schematic representation of a fully hydraulic tamping unit;

FIG. 3 shows a circuit diagram of a track geometry computer and a control device of the tamping machine; and

figure 4 shows a ballast bed acceptance record.

Detailed Description

Fig. 1 shows a tamping machine 38, C with a trailer 39, on which track-running chassis 34, 36 can run on a railway rail S. The tamping machines 38, C have a tamping unit 30 with a fully hydraulic drive and a measuring sensor 37, a lifting straightening unit 42, 43 for introducing a lifting force FH and a straightening force FR into the track, a work measuring system aw, bw, 35 and an acceptance register measuring system ar, br, 35. The task measurement systems aw, bw, 35 and the acceptance logger measurement systems ar, br, 35 are for example chord measurement systems. The trailer is coupled to the tamping machine by a tow bar 40. The tamping unit 30 has a standard opening width B of the tamping tool 29. The tamping machines 38, C also have a control system 19, a track geometry guiding computer 17 provided with a screen 20. Data is exchanged wirelessly with the infrastructure operator via the antenna 33. The work area is accurately detected in coordination by the GPS system 32.

Fig. 2 shows a tamping unit B with a fully hydraulic drive Z. The feed stroke 31 and the tamping force (via pressure sensors in the tamping cylinder hydraulic system) are detected by the sensors 23 and transmitted to the control computer 18, which forwards them to the track geometry computer 17 for processing. The braking deceleration of the tamping unit when it is inserted into the ballast bed is measured by an acceleration sensor bv. The harder the ballast bed, the higher the braking deceleration. The full hydraulic drive can adjust the opening width of the tamping arm 30 with the tamping tool 29 from the normal opening B to a larger width BE. In the position of damaged crushed ballast, ballast particles can therefore BE pushed from the sleeper box into the sleeper underneath in a tamping manner by a greater opening BE in order to supplement the partially damaged crushed ballast particles there with good ballast particles, thereby increasing the durability of the track position. The rail S is fixed to the sleeper 41.

Fig. 3 shows a circuit diagram of the track geometry computer 17 and the machine control device 19. The sensors of the fully hydraulic tamping units 18, 26 are read in and analyzed with the machine learning program ML. The mechanic knows the ballast bed condition through the screen 20 and can receive work instructions. At the end of the tamping operation, the track geometry computer 17 and the machine learning program ML create a ballast bed report 22 and a ballast bed record 21. These data are sent wirelessly 25 to a database of the infrastructure operator or machine owner or to the cloud. Ballast bed parameters under each tie are accurately detected by GPS and the tie assigned. The field position classified on track distance-km is assigned by means of a distance measuring wheel WMS.

Fig. 4 schematically shows a ballast bed record a. The recording column 1 shows the braking deceleration bv of the tamping unit, the recording column 2 shows the track height defect before operation, which has been derived from a preliminary measurement of the current track position and a comparison with the nominal track position, the recording column 3 shows the track bed hardness, and the recording column 4 shows the tamping force achieved. The record column 5 is an event column that indicates various special track conditions or track characteristics by markers 6, 7, 8, br. Symbol 6 denotes a rail joint, and symbol 7 marks a position in the track where ballast is damaged and therefore does not achieve a satisfactory tamping force. Symbol 8 represents a stored image, and Br indicates a bridge. The photos are embedded in the recording at individual single defects. If the mechanic activates them, the corresponding picture 8 is displayed. The individual defect points with damaged ballast 10 are shown, which can be seen on the one hand from the rapid reduction in the tamping force and on the other hand also from the reduction in the braking deceleration 11 of the tamping unit, since the ballast does not have a high immersion resistance in these points. 9 is a further fault location which occurs at the weld joint as visible at symbol 6. A machine learning program (or rule-based system) can detect and identify such individual defect locations relatively simply. If the course of variation of the height defect (entry bar 2) is compared with the course of variation of the hardness of the ballast bed (entry bar 3), it is found that they exhibit an approximately inverse ratio 12. A high point in height is created at the hard location. Generated in places where soft locations existSunken (concave). The degree of correlation between the two record fields can be determined by a correlation function. If the correlation is high, this affects the durability of the rail height position, since the ballast deformation has already correspondingly been produced to a greater extent. Hardness sigma of ballast bed _BH The higher the standard deviation of (a), the stronger the stiffness fluctuation and the greater the interaction force between the wheel and the rail and the lower the durability of the rail position. Whereas the average value of the ballast bed hardness 16, 17 indicates the degree of soiling wear of the ballast. The dirtier the ballast bed, the higher the ballast bed hardness 16, 17. The tamping force (the recording bar 4) varies in proportion to the hardness of the track bed. A very low value of the tamping force indicates a new layer 14 (new ballast) or a particular position 9, 10 with damaged ballast. Standard deviation sigma _V The lower the stiffness fluctuations and the better the durability of the track position. The horizontal line represents track kilometers (76400.).

An example of a gravel bed analysis report is shown below. A statistical evaluation is performed before this, which provides a general statement about the section processed. The analysis by the machine learning system ML provides statements about the durability of the rail position and the hardness of the ballast bed. If fault locations 9, 10 occur, their type, exact location, length and characteristic values are given. The transmission of these data to the infrastructure operator or responsible job master forms the pre-set basis for the job instructions for the next overhaul job. The analysis also provides an estimate of the rate of deterioration of the track position, which is critical to the time planning of the next major repair job. The data is also simply converted to a machine-readable form and transmitted.

Ballast bed report:

statistical evaluation

Durability of track position

The ballast bed has drawbacks. There is a low persistence of the track position.

The estimated result is a track position deterioration rate of 1.6 mm/year.

Hardness of ballast bed

The average ballast bed hardness was 254Nm.

The ballast bed is in a highly contaminated limit condition. It is recommended to clean the road bed. Critical fault locations (crushed/rounded ballast) have been found in the tamping area.

Replacement of ballast on 11 ties in the area 76580 is recommended.

Failure location 1

Type of location of failure	Start of	End up	Length of	Number of sleepers
					Minimum value	76578	76585	6.69m	11

Claims (5)

1. Method for automatically autonomously controlling a tamping machine (C) having a distance measuring device (WMS, GPS, 32) for precisely detecting the position of a working machine in a track for the construction of an overhead line, and having a signal detection device for actuators of a working unit (23, bv, 18, 26) of the tamping machine (C), characterized in that, during tamping, ballast bed data are detected by means of sensors (23, bv, 18, 26) and current ballast bed parameters are obtained therefrom and stored for a subsequent working process and analyzed by means of a machine learning device (17, ML), wherein an analysis of the ballast bed condition data (EF 7, S9, A3) is created on the basis of a machine learning technical method (ML, 17) and a plurality of working commands (EF 7, S9, A3) for an optimal working mode are determined therefrom and stored in consideration of the actual force disturbances occurring in the longitudinal direction of the track, wherein the tamping machine parameters and a plurality of working commands (EF 7, S9, A3) for the optimal working mode are determined and stored, wherein the working commands for the entire work process of the track and the tamping machine are automatically performed on the basis of the current working data and the subsequent working data.

2. Method according to claim 1, characterized in that the positionally accurate work orders (EF 7, S9, A3) for each tie area (41) to be tamped are transmitted to the control computer (17, 19) of the tamping machine (C) such that the tamping machine (C) approaches the respective longitudinal position in the track fully automatically as a function of the precisely detected coinciding longitudinal positions (WMS, rtk GPS, 32) in the track and the preset work order data (EF 7, S9, A3) and executes these work order data autonomously in the tie area (41) to be tamped as a function of the work order data (EF 7, S9, A3), thereafter automatically travels by means of the automatic travel system to the next tie area (41) to be processed as a function of the work order data (EF 7, S9, A3) and the work order data (EF 7, S9, A3) are executed and the next tie area (41) to be processed is approached until the predetermined work area is completed repeatedly.

3. Method according to claim 1, characterized in that preset work orders (EF 7, S9, A3) for each tie area (41) are displayed (20) to the operator and the operator adjusts and implements the preset work pattern.

4. A method according to any one of claims 2 or 3, characterized by creating a ballast bed condition record (fig. 4, a) during the work and creating a ballast bed condition report (fig. 4, a) by means of the machine learning device (17, ML) in connection with the results of the analysis of the ballast bed data and transmitting (33, 24, 23, 25) the ballast bed condition record (fig. 4, a) and the ballast bed condition report to an infrastructure operator.

5. Method according to any one of claims 1 to 4, characterized in that the analysis of the ballast bed data (A, fig. 4) by the machine learning device (17, ML) carried out together with the tamping work is transmitted to the operator for prompting the optimal work program.

Images